Preserving closedness of operators under summation (Q2512396): Difference between revisions

Given two unbounded linear operators \(A, B\) that are sectorial and resolvent commuting, an important question in operator theory is to establish whether or not the sum \(A+B\) is closed. Positive results to this problem are not only theoretically satisfying, but also have useful PDE applications, in particular, to obtain maximal regularity estimates for linear parabolic problems. The two most well-known results are due to \textit{G. Dore} and \textit{A. Venni} [Math. Z. 196, 189--201 (1987; Zbl 0615.47002)], and to \textit{N. J. Kalton} and \textit{L. Weis} [Math. Ann. 321, No. 2, 319--345 (2001; Zbl 0992.47005)]. Dore and Venni showed that a sufficient condition, in UMD Banach spaces, is that both operators have bounded imaginary powers (BIP). Kalton and Weis showed that, in a general Banach space \(X\), \(A+B\) is closed if one of the operators has a bounded holomorphic functional calculus (a condition stronger than BIP), while the other is R-sectorial (a condition weaker than BIP). The paper under review presents a version of this result, where the R-sectoriality is replaced by the following T-sectoriality (with angle \(\theta \in [0,\pi)\)) condition: \[ \begin{multlined} \forall n \in \mathbb{N} \;\forall r \in [e^{-1},1] \;\forall \phi \in [-\theta,\theta] \;\forall x_{0},\dots,x_{n} \in X\;\exists C>0 \;\exists a_{0},\dots,a_{n} \in \{f \in L^{\infty}(0,2\pi) : \|f\|_{\infty}\leq 1\}\\ \left\| t \mapsto \sum \limits _{k=0} ^{n} e^{ikt} (I+re^{-k+i\phi}A)^{-1}x_{k}\right\|_{L^{2}((0,2\pi);X)} \leq C \left\| t \mapsto \sum \limits _{k=0} ^{n} a_{k}(t)x_{k}\right\|_{L^{2}((0,2\pi);X)}.\end{multlined} \] This new condition is not directly compared to R-sectoriality, but Theorem 3.1, combined with \textit{L. Weis}' characterisation of maximal regularity [Math. Ann. 319, No. 4, 735--758 (2001; Zbl 0989.47025)] shows that, for generators of analytic semigroups on UMD spaces, the two notions coincide. In this context, the paper thus gives a new proof of Kalton-Weis' result. This new proof uses Kalton-Weis' ideas, but is based on direct estimates using the representation formula of \textit{G. da Prato} and \textit{P. Grisvard} [J. Math. Pures Appl. (9) 54, 305--387 (1975; Zbl 0315.47009)] for the inverse of \(\overline{A+B}\) rather than the more conceptual unconditionality lemma by Kalton and Weis [loc. cit., Lemma 4.1]. This makes the present version of the proof more technical than the original, but possibly more flexible.